S192
International Journal of Radiation Oncology Biology Physics
patients follow this trend may indicate patient-specific responses. The increase of mean HU in breast tissue was associated with the level of breast pain. Conclusion: Increase of mean HU in normal breast tissue was observed from the first to the last fraction CT sets acquired during APBI delivery. This increase is dependent on radiation dose. These changes in quantitative CT may reflect physical and/or chemical changes in the healthy tissue of the breast and may be used as an early predictor of acute and/or late treatment effects. Author Disclosure: X. Chen: None. C. Bergom: None. A.D. Currey: None. T.R. Kelly: None. C. Edwin: None. A. Montes: None. A. Li: None.
cancer, enhancing the possibility of using quantitative CT for early RT response assessment. Author Disclosure: G. Noid: None. A. Tai: None. Y. Liu: None. A. Li: None.
1066 Enhancement of Early Radiation Treatment Response Assessment by Mono-energetic Decomposition of Dual-Energy Computed Tomography G. Noid, A. Tai, Y. Liu, and A. Li; Medical College of Wisconsin, Milwaukee, WI Purpose/Objective(s): It has been shown that radiation can induce changes in CT textures (such as HU histogram) during radiation therapy (RT) and the change can potentially be used to assess RT response. This change will be amplified if X-Ray energy is reduced due to the rapid growth of the photoelectric effect below 100 keV. The purpose of this work is to investigate the use of low energy mono-energetic decompositions obtained from dual energy (DE) CT to enhance the detection of radiation induced changes in CT HU histogram. Materials/Methods: DECT data acquired for a CT phantom and for selected pancreatic cancer patients during routine CT-guided RT delivery using a CT scanner equipped with a sequential DE protocol was analyzed. The scanner rapidly performs two sequential acquisitions, the first at a tube voltage of 80 kV and the second at a tube voltage of 140 kV. Mono-energetic decompositions across a range of energies from 40 keV to 190 keV were reconstructed using an image-based material decomposition approach that uses mutual information between the two scans. The CT textures (e.g., HU histogram, mean HU) in several soft tissue inserts in the phantom and in the tumors (pancreatic heads) for the patients were measured across the spectrum of available mono-energies. The results from the mono-energetic decompositions were compared to those obtained from the standard protocol 120 kVp CT of the same subjects. Results: Data of mono-energetic decompositions of the CT phantom confirm the expected enhancement of soft tissue contrast as the energy is decreased. For instance in comparison to the 120 kVp scans, 40 keV decompositions demonstrated a 25 HU increase in the Liver tissue insert and a 50 HU decrease in the Adipose insert. For the pancreas patients, the low energy mono-energetic decompositions verified the increased change in CT HU histogram parameters during the delivery of RT. For the patients studied, the average reduction of the mean HU in tumor from the first to the last (the 28th) treatment fraction was 2.24 1.21 HU for the standard 120 kVp and 8.23 2.50 HU for the 40 keV mono-energetic decomposition. Conclusion: Low energy mono-energetic decompositions from DE CT substantially increase soft tissue contrast and increase the radiationinduced changes in CT HU histogram during RT delivery for pancreatic
1067 Prospective Validation of a Prognostic Computed TomographyeBased Radiomic Signature in Stage IV Non-Small Cell Lung Cancer E. de Jong,1 W. van Elmpt,2 R.T.H. Leijenaar,1 S. Carvalho,1 E.G.C. Troost,1,3 L.E.L. Hendriks,4 A.M.C. Dingemans,4 and P. Lambin2; 1 Department of Radiation Oncology (MAASTRO), GROW-School for Oncology and Developmental Biology, Maastricht University Medical Centre, Maastricht, Netherlands, 2Department of Radiation Oncology (MAASTRO), GROW e School for Oncology and Developmental Biology, Maastricht University Medical Centre (MUMC), Maastricht, the Netherlands, 3Institute of Radiooncology, Helmholtz-Zentrum DresdenRossendorf, Germany, Dresden, Germany, 4Department of Pulmonology, GROW-School for Oncology and Developmental Biology, Maastricht University Medical Centre, Maastricht, Netherlands Purpose/Objective(s): A recently described radiomic signature (Aerts et al; Nat Commun 2014), based exclusively on pre-treatment computed tomography (CT) imaging of the primary tumor volume, was found to be prognostic in independent cohorts of stage I-III lung and head and neck cancer patients. In this work, we investigated the prognostic value of this radiomic signature in a cohort of stage IV non-small cell lung cancer (NSCLC) patients. Materials/Methods: Patients were selected from a phase II study randomizing between nitroglycerin (NTG) patches or no additional treatment (NCT01171170). In total 169 of 223 included stage IV NSCLC patients could be retrieved. The prognostic index (PI) of the radiomic signature was determined using pretreatment CT scans for the complete cohort (CC) and for the subset of patients treated with Paclitaxel-Carboplatin-Bevacizumab with NTG (NTG) and without NTG (PCB). We tested the signature model fit in a Cox regression and assessed model discrimination with Harrell’s c-index. Patients were split in high and low signature prediction values by the median determined by Aerts et al. Kaplan-Meier survival curves between high and low signature predictions were compared with a log-rank test. The prognostic value of the model was tested for overall survival (OS) as well as for progression free survival (PFS). Results: After calibration, Kaplan-Meier survival curves were significantly different (P < 0.05) between high and low radiomic signature model predictions for OS for the CC (median OS low: 11.2, high: 6.7), for PFS for the CC (median PFS low: 5.9, high: 5.2) and for the PCB cohort (median PFS low: 7.2, high: 5.5). Conclusion: Overall, the radiomic signature was validated using CT images for stage IV NSCLC, demonstrating a moderate model fit and preservation of discrimination. Although the calibration slopes deviated from 1, which indicated an invalid relative risk model, the discriminative value of the radiomic signature was preserved. In the future we aim at improving the model by adding clinical information. Author Disclosure: E. de Jong: None. W. van Elmpt: None. R. Leijenaar: None. S. Carvalho: None. E. Troost: None. L. Hendriks: None. A. Dingemans: advisory board; Roche, Eli Lilly. P. Lambin: None.
Abstract 1067; Table 1 CC (n [ 169) Calibration slope on PI (P value) Harrell’s c-index (P value)
NTG (n [ 88)
PCB (n [ 81)
OS
PFS
OS
PFS
OS
PFS
0.596 (0.0080) 0.564 (0.017)
0.690 (0.054) 0.580 (0.0020)
0.438 (0.0026) 0.553 (0.15)
0.616 (0.064) 0.573 (0.033)
0.990 (0.97) 0.589 (0.023)
0.792 (0.42) 0.610 (0.0048)